1. Field of the Invention
The subject Invention is generally related to cameras and sensors and is specifically directed to a camera design that may use multiple imagers for day/night imaging, or for variable zoom. The design overcomes deficiencies of prior-art cameras wherein moving parts were required in the optical path, or wherein electronic zooming reduced image resolution. The cameras described may be either analog or digital.
2. Discussion of the Prior Art
Video cameras have become commonplace in modern life. Improvements in process technology have resulted in cameras offering high performance at steadily decreasing prices. While early cameras universally used vacuum-tube technology, modern cameras are solid-state devices using CMOS or CCD technology.
CCD and CMOS image sensors are typically organized into a planar array of photosensitive cells in orthogonal axes. When gated ON, each cell accumulates incident photons, yielding a net electrical charge in the cell. At the end of a defined exposure time, the accumulated charges are sequentially transferred from the exposed cells via switched-capacitor techniques.
Since the individual cells are merely accumulators of photons, they offer little or no inherent discrimination between various wavelengths of light. Indeed, such sensor arrays are often used for imaging using near-or-far infrared illumination. As a result of this broad-bandwidth behavior, such cameras are useful to produce monochrome video.
The overall dynamic range of such image sensors is limited. The top end of the dynamic range is limited by the cell's maximum-voltage capacity. That is, once the cell voltage has reached some specified maximum value (typically in the range of 1 to 5 volts), the cell is unable to retain any additional photon-induced charges.
The bottom end of the cell's dynamic range, which defines its sensitivity to low lighting conditions, is limited by imperfections in the array. There are two primary mechanisms. First, cell-to-cell nonuniformities exist in cell geometry, dark current, and sensitivity. These nonuniformities result in a fixed-noise pattern superimposed on images produced by the camera, even in the presence of a uniform illuminant. Second, thermal noise directly creates dynamic cell-to-cell voltage inequalities, which appear as ‘snow’ in rendered scenes. These two sources of noise limit the ultimate sensitivity of the system at low levels of illumination, since the effective signal-to-noise ratio decreases as illumination decreases.
A further reduction in sensitivity results if the array is configured to render color images. To render color images, the front surface of the array is overlaid with a color filter array such that adjacent cells are sensitive to different colors. A variety of such color filter schemes exist, including RGB, YCrCb, and Bayer patterns. While this enhancement does provide the array with a means to render color imagery, it reduces the overall sensitivity of the array. This reduction in optical sensitivity is due both to the non-ideal transmissivity of the filter to the desired wavelength, as well as to the elimination of energy at the remaining (non-desired) visible wavelengths.
Due to the ongoing advancement of digital techniques, a trend towards image sensors of ever-increasing resolution is evident. When the primary market for cameras was in television-related fields, there was not much demand for cameras to exceed the traditional 350 TV lines/480 scan lines common to broadcast television. Since then, a variety of video applications have emerged, driven by the emerging internet, digital still photography, and HDTV technologies. As a result, higher-resolution image sensors are in demand and are increasingly available. These higher-resolution imagers aggravate the problems of low-light sensitivity due to both their smaller cell geometries, and to the (typically) shorter exposure times they employ.
Many visual surveillance applications demand superior performance under very subdued illumination, such as night imaging with moonlight or even starlight as the ambient illuminant. While some progress has been made in improving the sensitivity of monochrome CCD imagers, such as the Watec 902H, the imagers remain insufficiently sensitive for moonlight or starlight operation. Conventional CCD or CMOS imagers are unable to produce usable video under these conditions. This problem has been remedied by the use of image intensifier technologies. These devices typically use a thin wafer of microscopically perforated opaque photoelectric material, biased by a strong electric field. Photons incident on the front of the plate produce an avalanche of photoelectrons through the microscopic pore, and energize a phosphor coating on the rear of the wafer. Currently available devices may exhibit luminous gains of perhaps 0.1 Foot-Lambert per foot-candle, which is enough to allow their use under starlight-illuminated conditions. Various commercial and military products use this technology, such as the military PVS-10 night vision scope, or the commercial OWL products produced by B.E. Meyer.
These image-intensifier devices have several operational limitations. First, the devices may be damaged if operated at ‘normal’ illuminations, such as daylight. Recent ‘gated’ and ‘filmed’ designs largely eliminate this problem. Also, these image-intensifier devices are not capable of rendering color scenes and are thus used in monochrome systems only.
As a result of the foregoing, separate cameras are generally required for operation under very low illuminations and for normal-to-low conditions. Prior-art systems not only use two image sensors; they use entirely separate optical paths including lenses. While this approach has utility, a significant weight and size penalty is incurred by the duplication of the lenses. This is particularly the case with night-vision cameras, which necessarily use very large lenses (160 mm and 310 mm lenses are commonplace) to increase the photon capture aperture of the system.
Another approach used in the prior art is the use of movable mirrors in the optical path. In this approach, a pair of mechanically linked mirrors is used to selectively pass the optical flux through one path, which includes an intensifier, or through a second path, which does not have in intensifier. While this approach has merit in its elimination of a redundant lens, it is not without drawbacks. One deficiency is in the difficulty of establishing precise physical positioning of the mirrors after they have been moved. This may be a serious deficiency if the system is used as a weapon sight. Another deficiency of a moving-mirror system is the mere presence of moving parts, which inevitably reduces the ruggedness and reliability of the system.
A related problem entails the need for image stabilization within the camera. Hand-held cameras are subject to the unavoidable vibration and tremors of the human hand. This problem is exacerbated when long-range telephoto lenses are employed; the viewed image is increasingly unstable as the magnification is increased. A variety of image stabilization techniques have been applied commercially. Typically, an orthogonal pair of gyroscopic sensors are disposed parallel to the image plane. The respective acceleration outputs are twice-integrated, then are used to correct the image position. An electronic technique, used in the Hitachi VM-H81 camcorder, offsets the image sensor array scanning clocks so as to compensate for instantaneous image displacement. An electromechanical approach is used in some lens adapters, manufactured by Canon. In this technique, a planar, optically transmissive object is placed in the optical path, and is mechanically articulated to displace the optical path as necessary to maintain a constant image position in the image plane. While useful, these approaches have their disadvantages. The optical method adds attenuation to the optical path, reducing the system's signal-to-noise ratio at low illumination levels. The electronic method necessarily sacrifices a bordered area at the top and sides of the image, sufficient to cover the peak angular excursions of the image. This effectively reduces the overall resolution of the rendered image.